How a Bridge Exploit Turned rsETH Into an Aave Solvency Test

The asset was not just rsETH

rsETH is meant to represent a claim connected to Ethereum staking and restaking activity. For a holder, its value depends on more than the token contract itself. It depends on reserves, redemption mechanics, liquidity venues, oracle assumptions and, crucially, the integrity of every bridge route capable of creating or transferring its representation across chains.

That last dependency became the weak point in April 2026. The Block reported that the roughly $292 million Kelp DAO incident affected the rsETH bridge and quickly raised bad debt concerns at Aave. The defining fact was not simply that tokens were stolen. It was that attackers could create rsETH that looked valid to downstream applications while lacking the economic backing users expected.

A bridge commonly works by locking or accounting for an asset on one chain, then authorizing an equivalent representation on another. The destination token is useful because it makes capital portable. It is dangerous because the recipient protocol must accept a message, proof or verifier set as evidence that the source side is properly collateralized.

If that evidence is forged or improperly authorized, the destination side can mint a claim without the corresponding locked asset. The token still has the right name, interface and transfer history. It may even initially trade close to its expected value. But it no longer represents the same economic claim.

This is the first lesson for lending markets: token identity is not collateral quality. Aave did not need an error in its lending code to face risk. It only needed to accept a token whose backing could fail somewhere outside Aave.

How the bridge problem entered Aave

The attack became a lending problem when the attacker deposited rsETH into Aave V3 and borrowed assets with deeper liquidity, including WETH. The Block reported that the exploit involved 116,500 rsETH and that the attacker used stolen rsETH as collateral to borrow substantial WETH, creating a risk of bad debt.

This is the crucial conversion. An attacker holding unbacked rsETH has a questionable claim. An attacker who posts it as collateral and borrows WETH has exchanged a questionable claim for a liquid, widely accepted asset. The exposure has shifted to depositors and the protocol’s safety mechanisms.

Aave’s automated logic evaluates a position using the value supplied by its oracle, the asset’s loan to value limit and its liquidation threshold. In simplified form:

Health factor = collateral value × liquidation threshold / debt value

A health factor above 1 means the position is generally not eligible for liquidation. A health factor below 1 permits liquidators to repay some debt in exchange for discounted collateral.

The problem was timing. If the rsETH oracle continued to report a price based on normal market conditions, the attacker’s position could appear healthy. Yet the token’s real economic support had deteriorated because the bridge had expanded supply without matching backing. A market price can lag that information, especially when transfers, redemptions or liquidity are impaired.

This was an economic correctness failure. The bridge, token and lending contracts could all follow their programmed rules. The system nonetheless produced an unacceptable result because the rules treated a failed claim as credible collateral long enough to extract real liquidity.

Why liquidation is not a magic solvent

Liquidation is often described as the mechanism that makes overcollateralized lending safe. That is incomplete. Liquidation only works if collateral can be sold or redeemed at a price sufficient to cover repaid debt, plus the liquidator’s incentive.

Assume a borrower posts collateral oracle valued at $100 million and borrows $75 million. If the position falls below its threshold, a liquidator may repay debt and seize collateral at a discount. But suppose the collateral’s realizable value has fallen to $55 million because backing is uncertain, withdrawals are paused or buyers have disappeared. The protocol cannot recover $75 million by selling an asset worth $55 million. The gap is bad debt.

A more practical measure is the liquidation gap:

Liquidation gap = outstanding debt - recoverable collateral proceeds

Recoverable proceeds are not the oracle price. They are the amount a liquidator can actually obtain after considering market depth, slippage, redemption delay, bridge restrictions and the liquidation bonus.

This distinction matters for assets linked to bridges. A price feed can remain technically correct as a measure of recent trading, while still failing to capture the speed at which redemption confidence is collapsing. Conversely, abruptly marking an asset to a distressed price can trigger liquidations that are impossible to execute at scale. Neither choice is painless.

Aave’s later response illustrated the unusual nature of the problem. Its April 28 recovery proposal temporarily adjusted the rsETH price oracle on Aave’s Ethereum Core and Arbitrum V3 markets to liquidate 89,567 unbacked rsETH securing 82,650 WETH and 821 wstETH in loans; the seized collateral was transferred to DeFi United’s Recovery Guardian before the oracle was restored. That was not routine market liquidation. It was an emergency intervention intended to turn impaired collateral into debt repayment.

Contagion came through shared assumptions

The failure path ran across several protocols:

1. A bridge accepted or relayed an invalid authorization.
2. Unbacked rsETH entered circulation.
3. Lending markets treated rsETH as eligible collateral.
4. Borrowed WETH and other assets left the pool.
5. The value of rsETH became contested.
6. Liquidation capacity depended on liquidity and recovery of backing.
7. Aave depositors and governance mechanisms faced the residual loss.

The correlation was hidden in plain sight. Aave may have had diversified borrowers and multiple collateral assets, but rsETH was exposed to a single external dependency: the integrity of its bridge and redemption backing. This is why nominal collateral diversification can overstate real resilience.

Risk managers should ask whether several assets share a bridge provider, verifier model, validator set, liquidity venue or redemption operator. If the answer is yes, they may all fail together during the same incident. Asset categories such as liquid staking token, stablecoin or wrapped asset are not enough. The underlying dependency graph matters more.

What risk controls can realistically do

The immediate tools are familiar: asset caps, borrowing caps, lower loan to value ratios, lower liquidation thresholds, supply freezes, borrowing freezes and isolation mode. Each reduces how much value can leave a lending pool before uncertainty is resolved.

But the Kelp DAO episode suggests these controls should be triggered by more than price volatility. A serious framework needs bridge-specific standards: verification decentralization, message security, minting limits, proof assumptions, redemption rights, chain-specific liquidity and emergency pause authority. It should also reassess assets whenever a material dependency changes, not only during a scheduled parameter review.

The proposed Aave framework reportedly adds checks for asset, bridge and chain risk, along with ongoing monitoring and a route to offboard assets that no longer meet the standard. That approach recognizes a governance reality often missed in DeFi: onboarding is not a one-time technical approval. It is a continuing decision to let an external risk sit inside a shared balance sheet.

For users, the question is similarly direct. Before looping a collateral asset or borrowing against it, ask what makes it redeemable, who can mint it, where its liquidity comes from and whether another protocol’s failure can invalidate its backing overnight.

The rsETH incident was not only a bridge exploit. It was a demonstration that DeFi composability turns infrastructure assumptions into credit risk. A protocol that lends against bridged collateral is not merely integrating a token. It is underwriting the bridge behind it.